intergovernmental grants and local government …...this paper considers the role of different...
TRANSCRIPT
1
Intergovernmental Grants and Local
Government Learning
Lyndal Keeton
Institutions and Political Economy Group (IPEG)
School of Economic and Business Sciences
University of the Witwatersrand
Private Bag X3, WITS 2050
Johannesburg, Republic of South Africa
+27.11.717.8065 (Tel.)
+27.11.717.8081 (Fax)
A work-in-progress from Lyndal Keeton’s PhD Dissertation supervised by Professor
Giampaolo Garzarelli for the Meeting of the ERSA Public Economics Workgroup
(16th and 17th May 2013)
Preliminary Draft 2nd
May 2013
Please do not quote, cite or redistribute without the author’s permission.
2
1. Introduction
The Public Economics literature on intergovernmental grants is extensive (e.g., Musgrave
1959; Bradford 1965; Bradford and Oates 1971; Oates 1972, 1999; Courant et al 1979;
King 1984; Carlson 1995; Oulasvirta 1997; Mieszkowski and Musgrave 1999; Volden
2007). In this extensive literature, grants are usually analysed according to consumer
behaviour theory. Under these analyses, the income and substitution effects determine
community spending (and ultimately community welfare). However, these effects shed
little light on how local governments can use grants to experiment with policy.
Casual empiricism shows that local governments routinely experiment with policy and
achieve varying degrees of success. One example is Mayor Bloomberg’s range of anti-
poverty experiments in New York City (Bosman 2009). Another example is that of
Mayor Mockus of Bogotá who introduced many innovative policy experiments. In one
experiment, Mayor Mockus employed mimes to direct traffic and, thereby, alleviate
traffic congestion (Caballero 2004). However, very little theory has been produced that
ties the policy experimentation issue with the role of intergovernmental grants. Hence, it
seems appropriate to start to develop models to understand how intergovernmental grants
can help local governments to perform policy experiments and learn.
This paper considers the role of different intergovernmental grants (e.g., a matching
grant versus a lump-sum grant) in local government policy experimentation. By taking a
different view of grants – namely, as incomplete contracts (Brennan and Pincus 1990,
Garzarelli 2006) – this paper suggests how certain grants can be instruments for the
creation and discovery of new knowledge (Garzarelli 2006). More precisely, this paper
3
intends to develop a model that tries to capture the learning effects that different types of
grants can engender.
2. Background
2.1 Laboratory Federalism
The theory of laboratory federalism suggests that such local government “laboratories”
are useful units in which to experiment. The reason for this is that local experiments that
prove unsuccessful will have a low cost to the federation as a whole, as the experiments
are only performed on a local level. Local experiments that prove successful will have a
high return as they can be applied by other local governments in the federation (Oates
1999: 1132). Furthermore, different experiments to solve the same problem can be run
simultaneously by different local governments to find the best policy solutions more
quickly than if experiments are tested in succession (Kollman, Miller and Page 2000:
102).
Put differently, a federation composed of local government “laboratories” is able to
solve problems by creating search networks. A local government with a particular
problem can find other local governments who can provide them with solutions because
the other local governments have already dealt with that particular problem. The
advantage of solving problems this way is that a federation is able to adapt to new
problems faster than a centralized government (Sabel 2004:12).
4
2.2 Grants as Incomplete Contracts
Garzarelli (2006: 249) suggests that intergovernmental grants can be used as a tool by a
central government to induce a local government to experiment with policy. In
submitting grants as a fiscal federalism counterpart to incomplete contracts1, Garzarelli
indicates that the type of grant communicates the donor’s preferred spending behaviour to
the recipient. Hence, a grant is an instrument of exchange between the donor and the
recipient.
However, there is no free lunch: with any type of grant comes the implication that
future grants will be related to how efficiently the current grant is used (Brennan and
Pincus 1990: 129; Garzarelli 2006: 248). Since the contract is incomplete, one or more
of the contracting parties may not be able to observe all information pertinent to the
contract. The rational grant recipient, therefore, has an incentive to use the funds
efficiently (or, at least, to the donor’s satisfaction) or else run the risk of not receiving
additional grants in the future.
Hence, an incomplete contract may only specify some conditions explicitly leaving the
remaining contract conditions implicit (See also, e.g., Seabright 1996). For this reason,
the grant can be viewed as a quid pro quo incomplete contract between the donor (the
central government) and the recipient (the local government) (Brennan and Pincus 1990;
Garzarelli 2006). This view of grants provides an institutional approach to grants that is
useful to understanding the role that grants can play in promoting experimentation and
learning in a federation.
1 An incomplete contract (also referred to as a relational contract) is a contract between two or more parties that
contains gaps or missing stipulations (Hart 1988:123). See, also, Baker, Gibbons and Murphy (2002) and Levin
(2003).
5
2.3 Grant Conditionality: Implicit and Explicit Conditions
Traditionally, there are two main classes of intergovernmental grants, namely conditional
and unconditional (see, e.g., King 1984). By viewing an intergovernmental grant as an
incomplete contract, a grant has conditions attached regardless of whether it is classed as
a conditional grant or not. Thus, we can distinguish between explicit grant conditions
(which give the grant its classification as a conditional grant) and implicit grant
conditions (Brennan and Pincus 1990).
An explicit grant condition is a formal stipulation made by the donor as to how the
grant should be spent (Brennan and Pincus 1990). It may specify which public good the
grant funds should be spent on, in the case of a conditional lump-sum grant, or the
matching ratio that the donor is committed to, in the case of a matching grant. Other
conditions may well be attached to the grant, though not explicitly stately, and thus are
implicit conditions of the grant. Hence, a grant donor can direct the spending of a grant
by using both the explicit conditions attached to the grant as well as the implicit
conditions attached to the grant.
Consequently, while an unconditional grant, e.g., an unconditional lump-sum grant,
does not have explicit conditions imposed, it may well have implicit conditions attached.
A conditional grant, e.g., a matching grant, is likely to have both the explicit conditions
imposed as well as additional implicit conditions attached. Thus, all grants, whether
explicitly conditional or not, are, in fact, conditional.
With different types of intergovernmental grants having different degree of
conditionality, it also means that different types of grants have different degrees of
6
leeway for a rational grant recipient to take advantage of (Garzarelli 2006: 249). Thus, to
reiterate, an unconditional lump-sum grant is likely to have relatively fewer conditions
attached as compared to a conditional matching grant, and vice versa. Thus, an
unconditional lump-sum grant has relatively more leeway as compared to a conditional
matching grant, and vice versa.
At this point we should note that the donor and the recipient have different interests as
regards how to spend grant funds. The donor or central government prefers to promote
those policies which it views as important, e.g., positive externalities and merit goods.
Thus, a donor will prefer a grant with explicit conditions attached, e.g., a matching grant.
The recipient or local government prefers to have leeway to spend the funds according to
local preferences. Thus, a recipient will prefer a grant with as few explicit conditions as
possible, e.g., an unconditional lump-sum grant (Oulasvirta 1997: 401).
2.4 Grant Conditionality: An example
An example of how local governments exploit the relative conditionality of grant can be
seen in the Canada Assistance Plan (CAP) grant scheme. This scheme offered federal
government assistance to welfare programmes administered at the provincial level.
Initially, the federal government contributed 50% of the costs of the welfare programmes
on an open-ended basis. In 1990 the scheme changed from a matching grant to a closed
matching grant for three of the ten Canadian provinces. This meant that federal shares to
welfare programmes in these three provinces could not grow by more than 5% per annum
from 1990 to 1994, relative to the 1989 welfare spending. If the provinces spend more
than 5%, they would not receive matching funds.
7
In studying the results of this change, (Baker, Payne and Smart 1999), find that overall
there was a decline in the growth of welfare expenditure in the three provinces. For the
most part, though, the provinces affected by the ceiling did not reduce the growth rates of
welfare benefits to correspond to the ceiling of the matching, as might be expected.
Instead the provinces rather concentrated on indirect methods to reduce spending. For
example, the three provinces implemented programmes to detect fraud, they reclassified
some welfare recipients as employable where previously they had been classified as
unemployable, and they changed incentives for beneficiaries to seek employment.
The responses of the three provinces were by no means identical, though. In attempts
to detect fraud, one province employed more fraud investigators while another started a
“snitch line.” One province reclassified single parents as employable when their
youngest child turned 12 years old. Some provinces reduced additional benefits, for
example, dental care and optometry, while some provinces changed the benefits allocated
to employed beneficiaries of the scheme. This is an example of a conditional matching
grant becoming more conditional when it’s changed to a conditional closed matching
grant. However, despite the increased conditionality, each of the three beneficiaries
responded differently, each exploiting the leeway within the grant to best satisfy local
preferences.
3. Intergovernmental Grants and Policy Experimentation
Similarly, intergovernmental grant can be a policy tool to provide the incentives for local
governments to experiment with policy. In this case, a policy is a single action taken by
the local government to resolve an issue of local public concern. To promote policy
8
experimentation at the level of local government, Strumpf (1998: 229) recommends that
fewer conditions be attached to an intergovernmental grant, it encourages greater
experimentation. This is taken one step further by Garzarelli (2006) who suggests the
rational grant recipient can take advantage of the leeway offered by an incomplete grant
contract by undertaking experiments, while keeping to any explicit conditions attached to
the intergovernmental grant. Hence, a grant that is less complete (e.g., an unconditional
lump-sum grant) has greater leeway for the recipient to engage in policy experimentation
than a grant that has relatively more complete (e.g., a matching grant). Thus, it is the
leeway attached to a grant that induces experimentation.
Furthermore, if it is the wish of the donor, e.g., the central government, to encourage
experimentation, it can reward a grant recipient, e.g., a local government, who engages in
experimentation with future grants.2 As mentioned, an intergovernmental grant is a
means of exchange. In this case, the donor’s future grants are rewarded by the donor in
exchange for experimentation with the current grant. Hence, intergovernmental grants
effectively ‘soften’ the budget constraint by providing a positive incentive for local
governments to use the grant effectively. In this case, the central government will use
intergovernmental grants to both encourage and reward experimentation.3
Since experimentation implies uncertainty for the local government in that it requires
the local government to implement policy choices which may or may not work. Strumpf
(2002) suggests that one reason for a local government not to engage in experimentation
is risk aversion. However, in addition to rewarding successful experiments,
2 Rewarding with future grants is not unlike the suggestion by Rose-Ackerman (1980) to offer prizes as a reward for
officials who experiment with policy. 3 The positive incentives created by soft budget constraints have not been examined in the literature. Typically, a soft
budget constraint is considered harmful to a federation as it creates perverse incentives for local governments to
overspend (e.g., Kornai 1986; Oates 2005; Vigneault 2007; Weingast 2009).
9
intergovernmental grants can be used to cushion unsuccessful experimentation.4 As
Kornai (1986:27) indicates, “a soft budget constraint has mercy on the loser," or, in this
case, mercy on the grant recipient whose experiment is unsuccessful. By rewarding the
action of experimentation, regardless of outcome, this can be achieved.
However, with any type of contract and not least of all with intergovernmental grants
comes the issue of "what can reasonably supposed to be observed and enforced" (Besley
and Persson 2011: 24). In this case, should the grant recipient choose not to engage in
experimentation, she will be held accountable by both the donor and her constituents.5
The donor can punish the grant recipient by increasing grants to other local
governments. The constituents of the local region will recognize that a portion of their
taxes is going to fund other regions and will hold the local government official
responsible (Goodspeed 2002: 418-419).6 Indeed, a case study of Norway's grant
reforms to replace conditional grants with unconditional lump-sum grants found that
local government officials had an incentive to "use spending decisions as strategic
instruments to achieve additional grants" because the public believed that the behaviour
of the officials could be the reason for cutbacks i.e. if cutbacks happened, the public
would hold the official responsible (Carlson 1995:56).
The objective of experimentation is to learn. And, thanks to policy experimentation,
learning – the absorption and generation of new knowledge (e.g., Arrow 1962) – can take
place at the local government level (Garzarelli 2006). A representative agent, such as a
local government representative, is attempting to find the best policy solution by
4 For local governments who abuse this soft budget constraint, it is possible for the central government levy a
punishment by increasing grants to other local governments. The constituents of the local government abusing the soft
budget constraint will recognize that a portion of their taxes is going to fund other regions (Goodspeed 2002: 418-419). 5 Such accountability will avoid problems of transfer dependence (Rodden 2003; Oates 2008).
6 Volden (1993: 73) finds that grant recipients do respond to the incentives created by intergovernmental grants but the
nature of response depends on institutions of recipient government.
10
undertaking the activity of experimenting with different policy choices. Since learning
occurs from both successful and unsuccessful experiments, it makes sense to reward all
experimentation rather than only successful experimentation. For the central government
it is less important whether the policy experiment is successful or not, what matters is the
overall growth of knowledge in a federation that is derived from multiple simultaneous
experiments.7
We have established that intergovernmental grants are incomplete contracts between
the donor and the recipient. We have also established that there is a role for
intergovernmental grants in laboratory federalism, namely in terms of providing the
incentives for local governments to engage in policy experiments. Because there are
many types of intergovernmental grants, particular types of grants are likely to have a
predisposition for recipient learning (Garzarelli 2006: 249). Now, if we consider how
intergovernmental grants can induce local government, it is possible to analyze how
different types of grants (e.g. a matching grant compared to a lump-sum grant) engender
different learning functions.
4. The Process of Learning
4.1 Mistake-ridden Learning
Learning may occur in all circumstances when agents have an imperfect
understanding of the world in which they operate - either due to lack of
information about it, or, more fundamentally, because of an imprecise
knowledge of its structure; or when they master only a limited repertoire
of actions in order to cope with whatever problem they face - as compared
to the set of actions that an omniscient observer would be able to conceive
7 For a parallel analysis in open-source software development see Garzarelli and Fontanella (2010).
11
of; or, finally, when they have only a blurred and changing understanding
of what their goals and preferences are. Dosi, Marengo and Fagiolo (2005:255)
Until now, we have considered the role that intergovernmental grants can play in
providing the incentives for grant recipients (in this instance, local governments) to
experiment and to learn. We now consider the process of experimentation that leads to
learning. As Arrow (1962:155) indicates, “learning can only take place through the
attempt to solve a problem and therefore only takes place during activity."
Learning is not a perfect process. Because experimentation and innovation involve a
degree of uncertainty, it is difficult to choose the correct policy straight away. Hence,
learning from experimentation implies some degree of mistakes (Nelson and Winter
1977:51). Hence, from now on we refer to the learning at the local government level as
mistake-ridden learning.
Dosi and Nelson (1994) suggest that learning is a process which is dynamic, i.e., it
occurs over time, and that it requires a selection mechanism. Selection is a "mechanism
providing differential rewards and penalties” (Dosi, Marengo and Fagiolo 2005: 318). In
the context of learning at the local government level, selection is the act of choosing the
most appropriate policy from the available policy options (Forte 1982: 234).
Each policy option under consideration for experimentation should have an associated
value or reward function attached to it that provides a guide to selecting the most
appropriate policy (Dosi and Nelson 1994: 154).8 Thus, the policy options are
competing with each other, and each policy option will be considered in terms of its value
relative to the values of the other policies.
8 This is similar to Lancaster’s (1966a) notion of a characteristic vector that determines the consumption of a particular
good.
12
In the case of local government learning, the local government official selects which
policy to apply by choosing that policy with the highest value, relative to the mean value
of all the available policy options, and switches between the policies depending on the
relative values for each policy (Dosi, Marengo and Fagiolo 2005). The act of switching
between the policy options is known as variation. Thus, learning at the local government
level takes place by the combination of selection and variation among the various policy
options.9
The selection of a policy requires feedback so that a local government official can
switch away from that policy if it proves unsuccessful, and, instead, experiment with
another policy. In this instance, the changes in the relative values of the different policies
provide this feedback. The speed of this feedback is important. Should the feedback
prove to be too slow, it can slow down learning. Similarly, if the feedback occurs too
quickly, it does not allow for experimentation and innovation, and can crowd out learning
(Dosi, Marengo and Fagiolo 2005: 319).
Notably the process of learning through selection and variation allows for multiple
equilibria. Hence, there can be multiple stable policy choices (Dosi and Nelson 1994:
154; Dosi, Marengo and Fagiolo 2005: 294). Thus, different local governments may find
different policy solutions for the same problem.
9 Note that pure selection implies no learning as it relies solely on random switching behaviour. Instead of
considering the value of each policy option, pure selection means that local government official will pick a policy
arbitrarily without taking into account past experience or whether the policy will be successful (Dosi, Marengo and
Fagiolo 2005: 318). Furthermore, there is a trade-off between selection and learning. Weak selection will allow slack
behaviour to persist. Instead, strong selection will hinder learning by crowding out the trial-and-error process
necessary for learning (Dosi, Marengo and Fagiolo 2005: 318-319).
13
4.2 Local Government Learning as a Process of Melioration
To understand how the grant recipient learns by experimenting with grant funding, we
consider the recipient’s behaviour upon receiving a grant. As Brennan and Pincus (1990:
136) note, “If the grant is in any way contingent on the recipient's behavior (either the
making of the grant at all or its size), then the grant tends to induce behavioral responses
in (potential) recipients." On receipt of a grant, using time and funds, the rational
behaviour of the local government representative is to take policy actions tied to the
grant, e.g., constructing a new road or reviving a local park. If the policy proves
successful, the recipient will continue with it. If this policy does not prove successful, the
recipient will switch to different policy.
One way to consider the process of experimentation and learning at the local
government level as if it is a process of melioration (see, e.g, Herrnstein and Prelec 1991;
Herrnstein et al. 1993; Metcalfe 2001).10
Melioration is the “the process of choosing that
alternative among the set of alternatives which currently has the higher yield in (value)"
(Herrnstein et al 1993: 150). Within the melioration process, at the local government
level, alternative policies are in competition with each other vying for the local
government representative’s time and funds.
It’s important to note that in the framework of melioration, it is the response or action
taken by the local government official that is the entity that we are interested in rather
than an individual (Herrnstein et al 1993: 177-180).
10
When analysing individual behaviour, whether maximization or, as in this case, melioration, we consider the
decision process to have taken place as though the individual has acted purposefully following this behaviour. As
(Buchanan 1982[1991]: 245) indicates: Individuals do not act so as to maximise utilities described in independently-existing functions. They confront
genuine choices and the sequence of decisions taken may be conceptualized, ex post (after the choices), in terms of
"as if" functions that are maximized. But these "as if" functions are, themselves, generated in the choosing
process, not separately from the process. (Buchanan 1982[1991]: 245).
14
Furthermore melioration is an alternative to the traditional utility maximization
framework. In a maximization framework, alternative policies are articulated together so
that they lose their individuality, i.e., only the utility attached to whole choice bundle
matters. Instead, melioration requires that the utility be unbundled and attached to each
individual policy so that each policy can be considered separately. This unbundled utility
is referred to as the value of the policy (Herrnstein and Pralec 1991: 154-155).
Melioration is appropriate to model experimentation and learning at the local
government level because it encompasses both the selection of a policy based on the
value attached to that policy and a process of switching to better policy choices.
Under the process of melioration, when a policy experiment is unsuccessful, the value
of the policy will decline relative to the values of other possible policies.
When the value of the policy is lower than that of an alternative policy, the local
government representative can switch to the policy which has the greater value.
Alternatively, though, if a policy is successful, the value of the policy will increase
relative to the values of other possible policies. Provided the value of the policy remains
higher than the values of alternative policies, the local government representative will
have no incentive to switch to a different policy. Thus, implementing a chosen policy
can have positive or negative effects on the relative value of the available policy options,
i.e. the values associated with the policy options change endogenously during
melioration. The process of melioration can continue up to the point where time and
funds are used up. Furthermore, the process will tend to or reach a meliorator's
equilibrium, which is where the average value of the alternative policies is equal
(Herrnstein and Pralec 1991).
15
To illustrate the process of melioration, consider how a local government representative
switches between two policies, 1x and 2x . For example, suppose the local government
needs to solve the problem of potholes on local roads. In this example, 1x may involve
simply fixing each pothole individually, while 2x may involve re-tarring the roads where
there are potholes. Each policy, ix , has a corresponding value function ( )i iv x which is the
amount or the value of satisfaction perceived to be obtained from ix . 1 1( )v x is the value
of 1x which is a faster solution with little inconvenience to drivers on the road while
2 2( )v x is the value of 2x that it is a long term solution because the potholes will not
reappear. The weighted mean of the values is ( ) ( )i i i
v x x v x=∑ . The value function
indicates relative preferences for different policies, so, for example,
1 1 2 2 1 2( ) ( )v x v x x x> ⇔ f , and, similarly, 1 1 2 2 1 2( ) ( )v x v x x x< ⇔ p . As mentioned, the
value function provides feedback to the local government representative as to the relative
success or failure of the policy chosen. Thus, the local government representative is
aware of both value functions and their mean. Accordingly, the local representative is
able to compare the two policies based on their relative values.
As previously indicated, selecting from the alternative policies can have positive
or negative effects on relative value of the policy. If the relative value of that activity
starts to decline, and falls below the relative value of another policy, the recipient will
switch to the policy with a higher value. Let the derivative of the value function
ii
i
vv
x
∂′=
∂for ease of notation. The sign of
iv′ indicates the whether the policy continues to
hold value, whether its value is appreciating or whether its value is diminishing, i.e.,
16
0iv′ < when the activity generates negative feedback (i.e. is an unsuccessful policy
experiment) and 0iv′ > when the activity ix generates positive feedback (i.e. is a
successful policy experiment).
Similarly, let 2
2
ii
i
vv
x
∂′′=
∂which is the rate at which the value of the policy increases or
decreases. In this instance of melioration, the changes in the relative values of the
different policies provide the feedback.
Thus, to sum up, local government learning can take place as a process of melioration
in which the local government representative selects a policy to implement by choosing
that policy with the highest value (relative to the mean value of all the available policy
options) and switches between the policies depending on the relative values for each
policy.11 Based on this process of selection (selecting a policy as an experiment) and
variation (switching between policies) learning by experimentation can take place.
5. Intergovernmental Grants: Leeway to Experiment
5.1 The Local Community Budget Constraint
Assume there are two levels of government: the higher-level, central government and the
lower-level, local government: the central government is the intergovernmental grant
donor and the local government is the intergovernmental grant recipient. More
specifically, within the local government, there is a local government representative who
11
Note that pure selection implies no learning as it relies solely on random switching behaviour. Instead of
considering the value of each policy option, pure selection means that local government official will pick a policy
arbitrarily without taking into account past experience or whether the policy will be successful (Dosi, Marengo and
Fagiolo 2005: 318).
17
is a representative agent of the local community acting according to the preferences of the
median voter.12 It is this local representative who determines how the grant will be spent.
The local government representative is aware of the local community’s budget
constraint which provides the limit or bound on local public spending. The budget
constraint incorporates both a monetary component and a time component. The
monetary component, initially, are local public funds which depend only on taxes raised
from the local community. We assume there is no local government borrowing. There is
a time component because, in this model, we are considering local government spending
on experimental policies. It takes time for the local government representative to
implement a policy experiment and to evaluate the experiment’s success or failure.
Assume that the local representative has a period of time or a budget term,T , in
which to spend local public funds, B . The assumption of both a monetary component
and a time component is in line with many public budgets allocated for a period of time.
This time period may be one year or, alternatively, over a longer period, such as three
years. This may be country dependent. Typically budgets are spent over one year but
countries such as Great Britain and South Africa have implemented medium-term
budgets, e.g., a term of three years, for public spending (Ajam and Aron 2007). Later we
will also see that learning at the local government level is a dynamic process i.e. learning
has a time component. Suppose, also, that the local representative can choose to spend
the local budget on two policies, 1x and 2x . Thus the budget constraint becomes:
1 1 1 1 2 2 2 2( , ) ( , ) ( , )c p t x c p t x Y B T+ = , 1
12
For selection of representatives see Besley and Coate (2003).
18
where ic is the cost of policy
ix , ic consists of both a time component
it and the price of
the policy ip where 1, 2i = . The total local community budget is Y which determined by
both local public funds, once again, B ; as well as the budget term, T , where, generally,
it T=∑ or, specifically in this case, 1 2t t T+ = .
We can rearrange the budget constraint equation 1 to provide the share of the
budget that is assigned to each policy, as follows:13
1 1 1 2 2 21 2
( , ) ( , )1
( , ) ( , )
c p t c p tx x
Y B T Y B T+ = . 2
Let
1 1 11
( , )
( , )
c p tz
Y B T= , and 3
2 2 22
( , )
( , )
c p tz
Y B T= , 4
such that iz is the share of income (expressed in terms of both time and funds) devoted to
policy ix . Thus, the local budget constraint can be expressed as
1 1 2 2 1z x z x+ = . 5
5.1 Melioration and Local Policy Choices
Suppose the local government receives a grant with which to experiment. The
melioration process is illustrated in Figure 1. Acting as meliorator, the local government
representative will always choose that policy which has the highest value of the two
policy choices, namely policy A and policy B. The values of the two policies, v(xA) and
13
See Metcalfe (2001).
19
v(xB) respectively, change endogenously as more of the policy is tried out. Figure 1,
thus, shows the how the value functions change relative to each other.
Consider Figure 1 Part A. Initially, Policy B has a higher value than Policy A.
However, the value of Policy B is declining and the value of Policy A is increasing, so at
ME1, the values of Policy A and Policy B are equal. Thus, at ME1, the local government
representative switches to Policy A. The value of Policy A continues to increase, but at a
decreasing rate, then falls while the value of Policy B begins to rise again. At ME2, the
values of both policies are equal, once again. Thus, ME2 is another switching point, and
the local government representative switches back to policy B, which has the higher
value beyond ME2.
This melioration process continues as long as there both time and funds are
available, i.e, until the grant funding runs out or until a time limit on spending the grant
funds is reached. The number of times the local government representative switches
between the available policies is an indication of the number of experiments undertaken,
and therefore, the extent of the learning that has taken place.
Figure 1 Part B shows a “learning path” along which policies will be implemented
based on the value functions. Throughout the melioration process, the better or more
successful policy (as determined by the value function) will be implemented.
20
Figure 1: Melioration between Two Policy Choices, A and B
The process of melioration reduces the choice set available to the policy
representative to only those policy choices where the relative value is higher than the
mean of the relative values. Thus, it is possible to remove some of the possible policy
choice mistakes, though not all.
V(xA)
ME3 ME2 ME1
Value of
Policy A
Value of
Policy B
V(xB)
V(xA)
ME3 ME2 ME1
Value of
Policy A
Value of
Policy B
V(xB)
Part A
Part B
21
Successful experiments, in effect, destabilize the system by innovating and
providing new policy solutions. These new policy solutions may even cause the budget
constraint to shift outwards. Unsuccessful policy experiments, in turn, provide stability
to the meliorating system by halting the progress of the unsuccessful experiments (Witt
1997: 490). The interaction between successful policy experiments and unsuccessful
policy experiments provides learning at the local government level.
5.2 Melioration and Local Government Experimentation as a
Replicator
The process of melioration can also be described by a simple dynamic replicator equation
(Loewenstein, 2010):14
( ) ( )( )i i i iz z v x v xγ= −& , 6
which shows that the change in the share of income direct to a policy ix over time,
iz& .
0iz ≥ is the share of income (once again, incorporating both time and local public funds)
directed to the policy ix , where 1,2i = . γ is the rate of switching between policies. The
process of melioration will take place provided 0γ > . Thus, γ determines the extent of
learning. ( ) ( )( )i iv x v x− is the difference between the value of the policy chosen
( )i iv x and the mean of the policy values ( )v x .
Thus, the share of income directed to
policy ix over time will increase provided 0γ > , 0iz > and ( ) ( ) 0i iv x v x− > or
( ) ( )i iv x v x> .
14 The dynamic replicator equation (1), which has been used in a variety of contexts, such as population genetics and
animal behaviour (REFS?).
22
Hence, the replicator equation (Equation 6) shows the following three equilibria (at
which melioration stops):
1. The rate of switching is zero, i.e., 0γ = ,
2. The share of income devoted to a particular good is zero i.e. 0=iz .
3. The value of current activity equals the average value of all activities i.e.
)()( iii xvxv = (which is the meliorator’s equilibrium).
The local government official will experiment by process of melioration as long at
there are both time and public funds available conduct policy experiments with (where
0iz i= ∀ ). Up to the point that the time and funds run out, the equilibrium that the
official’s meliorating behaviour will tend towards is the meliorator’s equilibrium, where
)()( iii xvxv = , i.e., all available policy choices have an equal value.
An optimal equilibrium point would be the case where equilibria (2) and (3) are
reached simultaneously for both policies, 1x and 2x . Consequently, this equilibrium
requires both that the share of the budget devoted to each policy is spent, i.e., 0=iz and
that the values of all policies tried out are equal and these values are equal also to the
mean value of all activities, i.e., ( ) ( )i i iv x v x i= ∀ .
If the share of the budget devoted to each policy is exhausted, the local
representative has used all available time and public funds to maximise the learning
potential from the grants given (by performing as many policy experiments as possible).
If the values of all policies tried out are equal, the local representative has reached the
meliorator’s equilibrium and both policies are valued equally.
Note that the process of learning by melioration allows for multiple equilibria.
Hence, there can be multiple stable policy choices (Dosi and Nelson 1994: 154; Dosi,
23
Marengo and Fagiolo 2005: 294). Here the objective is not to reach the highest point on
the budget constraint so that the local community consumes the most, but instead to learn
the most from the grant funds given. Unsuccessful policies can provide as much learning
as successful policy choices.
6. Types of Intergovernmental Grants and Learning
We can think of the local community budget constraint as a boundary of spending by the
local community determined by local public fund-raising efforts (local government
borrowing, local taxes etc). Similarly, we can consider the budget line that incorporates
an intergovernmental grant as another boundary, the shape of the boundary is dictated by
the type of grant. These two boundary lines delineate a corridor of potential local
government spending as a result of receiving the grant. Thus, we can think of the two
boundaries providing a corridor that can represent the propensity for the local
government to learn from policy experimentation that is funded by the grant. Hence, an
intergovernmental grant can open a learning corridor for policy experimentation.15
Here we consider the different learning functions attached to two commonly used
grants, namely a conditional matching grant and an unconditional lump-sum grant. The
reason for using these two types of grants is because one is explicitly conditional while
the other had no explicit conditions attached. A matching grant is considered conditional
under the traditional theory of intergovernmental grants. As discussed, it may have
further conditions attached (in the form of implicit conditions) rendering it even more
conditional than would seem on first glance. A lump-sum grant is considered
15
For more on corridor phenomena see Leijonhufvud (1973, 2009), Witt (1997), and Brenner (2002).
24
unconditional, however, as discussed, all grants have conditions attached and these
conditions can be implicit.16 Thus, the matching grant has greater conditionality than the
lump-sum grant. We assume that the grant donor encourages experimentation, and will
reward the activity of policy experimentation with future grants.
Consider the budget constraint (Equation 2), now with intergovernmental grants.
Suppose an unconditional lump-sum grant, G , is allocated to the local government.
Thus, the local budget constraint (and upper bound of the learning corridor) for the local
community with an unconditional lump-sum grant becomes:
1 1 1 2 2 21 2
( , ) ( , )1
( , ) ( , )
c p t c p tx x
Y B G T Y B G T+ =
+ +. 7
Therefore, in the case of an unconditional lump-sum grant, the income to both
goods increases. In comparison, if a conditional matching grant is allocated to the local
government at matching rate m , where 0 1m< < . The grant is conditional on local
spending on 1x only. Thus, the budget constraint (and upper bound of the learning
corridor) for the local community with conditional matching grant becomes:
1 1 1 2 2 21 2
(1 ) ( , ) ( , )1
( , ) ( , )
m c p t c p tx x
Y B T Y B T
−+ = . 8
Note that 0 (1 ) 1m< − < . Therefore, in the case of a conditional matching grant
income to only the grant-aided good (once again, in this case, 1x ) increases.
16
As Breton ([1985] 1987: 315) indicates, “Unconditionality, in turn, is the product of a reliance on a
theory of intergovernmental grants in which the decision makers are neither real governments, nor
competitive.”
25
6.1 Mathematical Representation
6.1.1 Lump-Sum Grant
Suppose that the central government provides the local government representative with a
lump-sum grant, G. Once again, the budget constraint becomes:
GbTxbtpxbtp +=+++ 222111 )()( 9
Or
1)(
)(
)(
)(2
221
11 =+
++
+
+x
GbT
btpx
GbT
btp 10
Where )(
)(
GbT
btp iii
+
+=α 11
Suppose that the central government provides the local government representative with a
lump-sum grant (G) with no explicit conditions attached. Thus G increases.
( )0
2
111 <+
+−=
GbT
btp
dG
dα 12
( )0
2
222 <+
+−=
GbT
btp
dG
dα 13
Once again, proceeding along the lines of Metcalfe (2001), the melioration equilibrium
can be specified with matching conditions, which for the matching grant are:
)()( 2211 xvxv = 14
12211 =+ xx αα 15
Thus using equations (12) and (13) the total derivative of equation (10)
22112211 αααα dxdxdxdx −−=+ 16
becomes
( ) ( )( )
dGGbT
xbtpxbtpdxdx
2
2221112211
+
+++=+αα
Substituting equation (9)
26
dGGbT
dxdx+
=+1
2211 αα
The system of adjustments that result from the lump-sum grant can be represented as
follows:
1 2 1
1 2 2
0
1v v dx
dx dGbT G
α α
′ ′− = +
Solving yields:
2 1 1 2 0v vα α′ ′∆ = + > 17
1 2 0( )
dx v
dG bT G
′= >
∆ + 18
2 1 0( )
dx v
dG bT G
′= >
∆ + 19
Hence, an increase in G from a lump-sum grant results in an increase in policy
experimentation on both goods and, consequently, learning takes place. The relative
values of the goods remain as per the median voter’s preferences. Note that decision to
choose between policy 1, x1, or policy 2, x2, depends on the change in the relative values
of the two policies, v1 and v2.
Now iii xz α= ,
( )1 2 2 1 1
1 2
dz v x v x
dG bT Gα α
′ ′−= ∆ + 20
( )2 1 1 2 2 1
1 2
dz v x v x dz
dG bT G dGα α
′ ′−= = − ∆ + 21
27
With a lump sum grant there is an increase in the funds to both policy experiments and
the share of funds allocated to each is determined by the relative value of the policy
experiment iv and the amount of each experiment undertaken ix .
6.1.2 Matching Grant
Suppose that the central government provides the local government representative with a
matching grant with the explicit condition that it be spent on 1x , i.e. the matching grant
increases the spending on 1x by making it relatively cheaper.17
Let the matching grant
rate be m where 10 <≤ m , so, for example, if the central government decides to give 50%
of what the local government has spent on the public good to them as a grant then
5.0=m . Thus the budget constraint becomes:
bTxbtpxbtpm =+++− 222111 )())(1( 22
Or
1 1 2 21 2
(1 )( ) ( )1
( ) ( )
m p bt p btx x
bT bT
− + ++ = 23
Where 1 1
1
(1 )( )
( )
m p bt
bTα
− += 24
And 2 2
2
( )
( )
p bt
bTα
+= 25
Thus:
0111 <+
−=bT
btp
dm
dα 26
02 =dm
dα 27
17
Thus the matching grant changes the relative prices, making the public or grant-aided good relatively cheaper.
28
The matching grant will change 1α only. Proceeding along the lines of Metcalfe (2001),
the melioration equilibrium can be specified with matching conditions, which for the
matching grant are:
)()( 2211 xvxv = 28
12211 =+ xx αα 29
Then the system of adjustments that result from a matching grant can be represented as
follows:
1 2 1
1 2 2 1 1
0v v dx
dx x dα α α
′ ′− = −
Solving yields:
2 1 1 2 0v vα α′ ′∆ = + > 30
1 2 1
1
0dx v x
dα
′= − <
∆ 31
2 1 1
1
0dx v x
dα
′= − <
∆ 32
Now iii xz α= ,
1 1 1 2 11 1
1 1
0dz dx v x
xd d
αα
α α
′= + = >
∆ 33
2 2 2 1 12
1 1
0dz dx v x
d d
αα
α α
′= = − <
∆ 34
A matching grant in favour of x1 over x2 increases the experimentation on x1 at the
expense of x2.
Thus using equations (31) and (32) the total derivative of the budget constraint (equation
21)
29
22112211 αααα dxdxdxdx −−=+ 35
becomes
( )dmxdm
bT
btpxdxdx 11
1112211 ααα =
+=+ 36
Then the system of adjustments that result from the matching grant can be represented as
follows:
=
′′
dmxdx
dxvv
112
1
21
210
ααα
Solving yields:
02112 >′
+′
=∆ vv αα 37
02111 >∆
′=
vx
dm
dx α 38
01112 >∆
′=
vx
dm
dx α 39
Equations (36) and (37) are both positive. This means that a matching grant to one good
can increase experimentation on both policies in absolute terms. However, the change in
expenditure on both policies is dependent on the grant-aided good, x1. This means that
although experimentation overall increases, the experimentation is constrained by the
commitment of the grant-recipient to honour the explicit matching condition of this grant.
By comparing equations (36) and (37) with the parallel equations for a lump-sum grant,
equations (16) and (17), the system of adjustments shows that under a lump-sum grant,
the grant recipient does not face the same constraint. Hence, a lump-sum grant provides
the grant recipient with greater leeway to experiment than the matching grant.
Now iii xz α= ,
0)( 111121
1
11 <
∆
+′
−=∂
∂=
bT
btpxv
md
dz
dm
dz αα
α 40
30
0)( 111121
1
22 >∆
+′
=∂
∂=
bT
btpxv
md
dz
dm
dz αα
α 41
Note that in equations (38 and 39), the change in zi is dependent only on the value of x1.
Thus, the matching grant is yields a different outcome to the lump-sum grant. As
mentioned, the lump-sum grant increases funds to both policies depending on the relative
values of the experiments. With a matching grant, however, the funds spent on each
policy are, once again, constrained by the expenditure on the grant-aided good x1.
6.2 Graphical Representation
The process of learning by melioration can be represented graphically in Figure 3,18
although the graph is not a strict representation thereof. Assume, for illustrative
purposes, that time is constant.
Figure 2: Graphical Representation: Learning under Lump-sum Grant
compared to Matching Grant
18
From Herrnstein’s (1974) ‘matching law,’ since xi represent responses and if we consider that the value functions
v(xi) act as reinforcements, then ( )
1 1
( ) ( )1 2 1 2
x v x
x x v x v x=
+ +, and, hence,
( )1 1
( )2 2
x v x
x v x= .
31
In Figure 3, the matching grant-aided policy, 1x , is on the horizontal axis and an
alternative policy,2x , on the vertical axis, along with their relative values. The budget
constraint in the absence of intergovernmental grants (Equation 2) is represented by AC.
BD represents the shift in the budget constraint due to the lump-sum grant (Equation 7).
AK represents the shift in the budget constraint due to the matching grant which specifies
that funds spent on the grant-aided policy, 1x (Equation 8). The meliorator’s equilibrium,
as discussed previously, occurs when the value functions for both policies are equal,
1 1 2 2( ) ( ) ( )iv x v x v x= = . In Figure 3, the meliorator’s equilibrium is represented by a
45°degree line from the origin. To the left of the 45° line, the alternative policy, 2x , has a
greater value than the grant-aided policy, 2x . Thus, to the left of the 45° line,
2 2 1 1( ) ( )v x v x> and, hence, 2 1x xf . Similarly, to the right of the 45° line, the alternative
45°
x2,v2(x2)
alternative policy,
value of x2
B
A
x1, v1(x1)
grant aided policy,
value of x1
D C
J
Equilibrium
F
E
K
32
policy, 2x , has a lower value than the grant-aided policy,
2x . Thus, to the left of the 45°
line, 2 2 1 1( ) ( )v x v x< and, hence, 2 1x xp .
First, consider Figure 3 for the unconditional lump-sum grant (once again, BD).
The local representative receives this unconditional lump-sum grant from the central
government donor. Then, behaving as a meliorator, the local government representative
uses the time and funds from the grant to experiment the two different policies available,
namely 1x and 2x . As shown in Figure 3, the local representative will start with the
policy with the higher value, and continue to implement it until the value of the policy
declines and the other policy has a higher value.19
At this point, the representative will
switch to the other policy. This process will continue until the representative reaches the
upper bound or the budget constraint BD. The potential for switching between policies is
represented by the area of the corridor, ABCD.
Secondly, in parallel, consider Figure 3 for the conditional matching grant (once
again, BD). The local representative receives this conditional matching grant from the
central government donor. Then, behaving as a meliorator, the local government
representative uses the time and funds from the grant to experiment the two different
policies available, namely 1x and
2x . As shown in Figure 3, the local representative will
start with the policy with the higher value, and continue to implement it until the value of
the policy declines and the other policy has a higher value. At this point, the
representative will switch to the other policy. This process will continue until the
representative reaches the upper bound or the budget constraint AK. The potential for
switching between policies is represented by the area of the corridor, ACDJ.
19
As Breton ([1985] 1987: 315) indicates, “unconditional money…will tend towards areas where
competition is greater.”
33
The optimal point under an unconditional lump-sum grant will be F where the
values of the two activities are equal and the public budget exhausted, i.e. the recipient
has made the most of the funds available. However, as the graph is shown in Figure 3,
point F is unattainable under a conditional matching grant of the same size as the
unconditional lump-sum grant. Thus, the optimal point under the conditional matching
grant will be E, which is inferior to F, the optimal point under the unconditional lump-
sum grant. Note that it is possible for another matching grant to make point F attainable
(see Appendix 1).
However, the point of this type of policy experimentation is learning. As
mentioned, the potential for switching between policies is represented by the area of the
corridor created by the intergovernmental grant. Recall from equation 6, that the capacity
for learning from melioration is γ . Thus, the capacity for learning attached to an
intergovernmental grant, γ , is proportional to the area of the learning corridor bounded
by the local community budget constraint prior to grants and the local community budget
constraint after the grant is awarded. The learning corridor attached to the unconditional
lump-sum grant is area ABCD. In comparison, the learning corridor attached to the
conditional matching grant is area ACDJ. Thus, the potential learning from
experimentation attached to the unconditional lump-sum grant ( Gγ ) is proportional to
area ABCD, i.e., G ABCDγ ∝ . In comparison, the potential learning from
experimentation attached to the conditional matching grant ( mγ ) is proportional to area
ACDJ, i.e., m ACDJγ ∝ . The difference in areas is area ABJ which indicates that a
grant recipient may learn less from a conditional matching grant than an unconditional
lump-sum grant, i.e., m Gγ γ< .
34
Finally, area ABJ represents the potential for learning that is crowded out by the
conditions imposed by the conditional matching grant. Thus, difference between the
learning capacity of an unconditional lump-sum grant and the learning capacity of a
conditional matching grant is proportional to ABJ, i.e., G m ABJγ γ− ∝ . Thus, this shows
that a local government can learn more from an unconditional lump-sum grant than a
conditional grant.
7. Conclusion and Policy Implications
This paper considers a model of learning by a process of melioration that can illustrate
how different types of intergovernmental grants can induce different degrees of learning
at the local government level.
This paper additionally provides a positive role for a soft budget constraint at the
local government level. Provided the local governments are accountable to both the
central government donor and their local constituents, a soft budget constraint can
stimulate experimentation and learning.
The flypaper effect is the phenomenon in which an intergovernmental grant leads to
greater public spending than an equivalent increase in community income, e.g. a national
tax cut (Bailey and Connolly 1998: 335).20 It’s dubbed the flypaper effect because money
sticks where it hits.21
20
The flypaper effect contradicts "veil hypothesis" which suggests that an unconditional lump-sum grant will have the
same effect as a decrease in taxes of the same amount: the unconditional grant is a veil for a federal tax cut (Oates
1999). 21
The term “flypaper effect” was coined by Arthur Okun (Inman 2008:1).
35
The flypaper effect can be broken down into macro effects and micro effects
(Brennan and Pincus 1990). The macro effect is the effect of a grant on local
government expenditure. This is the conventional form of the flypaper effect, as
described. Here there is a choice between whether the grant will be spent on private
consumption or on public consumption. When a lump-sum grant is given, the spending is
skewed towards public goods. The micro effect is less conventional and relates to the
different effects that different grants have on local government spending. One example
of this is that a conditional lump-sum grant is typically more stimulating for that public
good for which it is intended than a conditional matching grant or an unconditional grant
(Brennan and Pincus 1990: 134-136).
While previously, and indeed in the first generation theory of fiscal federalism, the
flypaper effect was viewed as an anomaly (Oates 2008: 324-325), much of the literature
(e.g., Courant, Gramlich and Rubinfeld 1979; Brennan and Pincus 1990; Aronson and
Munley 1996; Volden 1999; Knight 2002; Volden 2007; Oulasvirta 1997; Inman 2008) is
in agreement that the flypaper effect is an inevitable outcome of intergovernmental
relations.22
As Inman (2008: 9-10) notes, "once viewed as an anomaly, the flypaper
effect should now be seen as a reality of fiscal politics, and its study as an opportunity to
fashion central government transfer policies and intergovernmental fiscal institutions that
better reflect citizen preferences for public goods."
This paper shows that money will stick where it hits, when the policy experiment is
successful. When the policy experiment is unsuccessful, though, the funds will shift
towards experiments that appear to have greater potential for success.
22
For a review of the alternative explanations for the flypaper effect given in the literature see Bailey and Connelly
(2008).
36
37
8. References (incomplete)
Ajam, Tania and Aron, Janine 2007. “Fiscal Renaissance in a Democratic South Africa,”
CSAE Working Paper 2007-10.
Aronson, J. Richard and Vincent G. Munley 1996. “(Non)equivalence in a federalism:
Dual tax shares, flypaper effects and Leviathan,” Public Choice 89(1-2): 53-62
(October).
Arrow, Kenneth. E. 1962. “The Economic Implications of Learning by Doing,” Review of
Economic Studies 29(3): 155-173(June).
Atkinson, Anthony B. & Stiglitz, Joseph. E. 1969. “A New View of Technological
Change,” Economic Journal 79(315): 573-578(September).
Bailey, Stephen J. and Stephen Connolly 1998. “The Flypaper Effect: Identifying Areas
for Further Research,” Public Choice 95(3/4): 335-361 (June).
Baker, George, Robert Gibbons, and Kevin J. Murphy 2002. “Relational Contracts And
The Theory Of The Firm,” Quarterly Journal of Economics 117(1): 39-84 (February).
Baker, Michael, A. Abigail Payne, and Michael Smart 1999. “An Empirical Study of
Matching Grants: the ‘cap on CAP’” Journal of Public Economics 72(2): 269-288
(May).
Besley, Timothy and Stephen Coate 2003. “Centralized versus Decentralized Provision of
Local Public Goods: A Political Economy Approach,” Journal of Public Economics
87: 2611-2637 (December).
Besley, Timothy and Torsten Persson 2011. “Fragile States and Development Policy,”
Working Paper EOPP/2011/22 LSE Sunatory and Toyota International Centres for
Economics and Related Disciplines. Available online:
http://sticerd.lse.ac.uk/dps/eopp/eopp22.pdf Last Accessed 25 April 2013.
Bosman, Julie 2009. (April 22). “City Ends Some Antipoverty Experiments,” New York
Times Available online: http://cityroom.blogs.nytimes.com/2009/04/22/bloomberg-
ends-some-antipoverty-experiments Last Accessed 22nd March 2013.
Bradford, David F. and Wallace E. Oates 1971. “Towards a Predictive Theory of
Intergovernmental Grants,” American Economic Review 61(2): 440-448(May).
Brennan, Geoffrey and Jonathan Pincus 1990. “An Implicit Contract Theory of
Intergovernmental Grants,” Publius: The Journal of Federalism 29(4): 129-44(Fall).
Brenner, Thomas 2002. “A Behavioural Learning Approach to the Dynamics of Prices,”
Computational Economics 19(1):27-94 (February).
Breton, Albert 1987. “Towards a Theory of Competitive Federalism,” European Journal
of Political Economy 3 (1-2): 263-329.
Buchanan, James M. 1982[1991]. “Order Defined in the Process of its Emergence,”
Literature of Liberty 5(5): 7-58 (Summer).
Buchanan, James M. and Gordon Tullock 1962. The Calculus of Consent: Logical
Foundations of Constitutional Democracy. Ann Arbor: The University of Michigan
Press.
Courant, Paul N., Edward M. Gramlich and Daniel L. Rubinfeld 1979. “The Stimulative
Effects of Intergovernmental Grants: Or Why Money Sticks Where it Hits,” Chapter 2
in Peter Mieszkowski and William H.Oakland (Eds) Federalism and Grants-in-Aid: 5-
22 Washington D.C.: The Urban Institute.
38
Demsetz, Harold 1969. “Information and Efficiency: Another Viewpoint,” Journal of
Law and Economics 12(1): 1-22(April).
Dosi, Giovanni, Luigi Marengo and Giorgio Fagiolo 2005. “Learning in Evolutionary
Environments,” Chapter 9 in Kurt Dopfer (Ed.) The Evolutionary Foundations of
Economics: 255-338. Cambridge: Cambridge University Press.
Forte, Francesco 1982. “The Law of Selection in the Public Economy as compared to the
Market Economy,” Public Finance = Finances publiques 37(2): 224-245.
Garzarelli, Giampaolo 2006. “Cognition, Incentives, and Public Governance: Laboratory
Federalism from the Organizational Viewpoint,” Public Finance Review 34(3): 235-
257(May).
Goodspeed, Timothy J. 2002. “Bailouts in a Federation,” International Tax and Public
Finance 9:409-421(MM).
Hart, Oliver 1988. Firms, Contracts, and Financial Structure Oxford: Oxford University
Press.
Herrnstein, Richard J. and Drazen Prelec 1991. “Melioration: A Theory of Distributed
Choice,” Journal of Economic Perspectives 5(3): 137-156(Summer).
Herrnstein, Richard J., George F. Loewenstein, Drazen Prelec and William Vaughan Jr.
1993. “Utility Maximization and Melioration: Internalities in Individual Choice,”
Journal of Behavioral Decision Making 6(3):149-185(September).
Inman, Robert 2003. “Transfers and Bailouts: Enforcing Local Fiscal Discipline with
Lessons from U.S. Federalism,” Chapter XX in Jonathan Rodden, Gunnar Eskelund
and Jennie Litvack (Eds.) Fiscal Decentralization and the Challenge of Hard Budget
Constraints: 35-83 Cambridge, MA: MIT Press.
Inman, Robert 2008. “The Flypaper Effect,” NBER Working Paper Series. Working
Paper 14579. http://www.nber.org/.
Kollman, Ken, John H. Miller, and Scott E. Page 2000. “Decentralization and the search
for policy solutions,” Journal of Law, Economics, and Organization 16(1)102-128
(April).
Kornai, János 1986. “The Soft Budget Constraint,” Kyklos 39(1): 3-30.
Knight, Brian 2002. “Endogenous Federal Grants and Crowd-out of State Government
Spending: Theory and Evidence from the Federal Highway Aid Program,” American
Economic Review 92(1): 71-92 (March).
Lancaster, Kelvin 1966a. “Change and Innovation in the Technology of Consumption,”
American Economic Review 56(1/2): 14-23 (March).
Lancaster, Kelvin J. 1966b. “A New Approach to Consumer Theory,” Journal of
Political Economy 74(2): 132-157 (April).
Leijonhufvud, Axel 1973. “Effective Demand Failures,” Swedish Journal of Economics
75(1): 27-48 (March).
Leijonhufvud, Axel 2009. “Macroeconomics and the Crisis: A Personal Appraisal,”
Centre for Economic Policy Research (CEPR) Policy Insight No. 41 (November)
Available online: http://www.cepr.org/pubs/policyinsights/PolicyInsight41.pdf Last
Accessed 25 April 2013.
Levin, Jonathan 2003. “Relational Incentive Contracts,” American Economic Review
93(3): 835-857 (June).
Metcalfe, John Stanley 2001. “Consumption, Preferences and the Evolutionary Agenda,”
Journal of Evolutionary Economics 11(1): 37-58(January).
39
Oates, Wallace E. 1999. “An Essay on Fiscal Federalism,” Journal of Economic
Literature 37(3):1120- 49(September).
Oates, Wallace E. 2005. “Toward a Second-generation Theory of Fiscal Federalism,”
International Tax and Public Finance 12(4): 349-373(August).
Oulasvirta, Lasse 1997. “Real and Perceived Effects of Changing the Grant System from
Specific to General Grants,” Public Choice 91: 397-416
Rose-Ackerman Susan 1980. “Risk Taking and Reelection: Does Federalism Promote
Innovaton?” Journal of Legal Studies 9(3): 593-616 (June).
Sabel, Charles 2004. “Beyond Principal-Agent Governance: Experimentalist
Organizations, Learning and Accountability,” in Ewald Engelen and Monika Sie
Dhian Ho (eds.) De Staat van de Democratie. Democratie voorbij de Staat
Amsterdam: Amsterdam University Press, (2004). Available online:
http://www3.law.columbia.edu/sabel/papers/Sabel.definitief.pdf Last Accessed 24
April 2013.
Seabright, Paul 1996. “Accountability and Decentralisation in Government: An
Incomplete Contracts Model,” European Economic Review 40:61-89.
Strumpf, Koleman S. 1998. “A Predictive Index for the Flypaper Effect,” Journal of
Public Economics 69(3): 389-412 (September).
Strumpf, Koleman S. 2002. “Does Government Decentralization Increase Policy
Innovation,” Journal of Public Economic Theory 4(2): 207-241 (April).
Vigneault, Marianne 2007. “Grants and Soft Budget Constraints,” Chapter 5 in Robin
Broadway and Anwar Shah (Eds.) Intergovernmental Fiscal Transfers: Principles and
Practice: 133-171 Washington D.C.: The World Bank.
Volden, Craig 1999. “Asymmetric effects of intergovernmental grants: Analysis and
implications for US welfare policy,” Publius 29(3): 51-73 (Summer).
Wildasin, David E. 2004. “The Institutions of Federalism: Toward an Analytical
Framework,” National Tax Journal 57(2): 247-272 (June).
Witt (1997)
Wohlgemuth, Michael 2002. “Evolutionary Approaches to Politics,” Kyklos 55(2): 223-
246
40
Appendix 1
x1 value of the grant preferred policy
45°
V2 value of
the alternative
policy
B
A
D C
J
Meliorator’s Equilibrium
F
E
K
x1 value of the grant preferred policy
45°
B
A
D C
Meliorator’s Equilibrium
F
E
K
V2 value of
the alternative
policy
Matching Grant 1
Matching Grant 2
41
Matching Grant 3
45°
B
A
J Meliorator’s Equilibrium
F
E
V2 value of
the alternative
policy
C D x1
value of the grant preferred policy
K